On Tuesday 27 September 2005 08:02 pm, Spehro Pefhany scribbled:
> Hi, Neil:-
>
> I often triplicate critical data, but I like to separate it into different
> pages, and avoid certain locations. You do have to allow for 3 write
> cycles, which is quite a few msec worst case with Fowler-Nordheim tunneling
> NV memory, not a problem at all if you use FRAM (but it's single-sourced
> and more expensive). I also attempt to repair bad data.

I had looked into FRAM recently when someone posted another thread asking for 
an ultra-fast EEPROM.  Seems really nice.  Yes, more expensive, but overall 
cost increase for the whole project is negligible.  The single-source does 
bother me, but if it's fully compatible with regular serial EEPROMs, then I 
shouldn't be too worried.  More importantly though is if they really work as 
advertised.  IIRC the datasheet said something like "instant" writes, but in 
some other doc I had seen something about 72us write time, which is 
noticeable.  I'm not so much interested in lowering the write cycle time as I 
would be in eliminating it, since eliminating it would let me remove all the 
existing code that queues the datasets/address to be written and farms it out 
to the EE write code as the EE device becomes free.  Either way, I'm waiting 
on FRAM samples so I can try it out.


> It depends a bit on what you are worried about as a cause of data
> corruption, of course-- whether it's EEPROM failure or some external
> influence.

Everything -- EE failure, power failure, master working-data corruption, my 
own code bugs, EE getting eaten by a dog ... anything. :-)

What I did was to triplicate the data, with each instance of any dataset 
written to different pages.  On startup/readback, any matching pair is 
considered valid.  I wasn't sure if EEPROMs could fail in a specific way -- 
such as all bits getting reset to 1's, etc, so I modified the data in each of 
the 3 groups -- currently, group 1 is raw, group 2 is bit-inverted, and group 
3 has alternating bits inverted.  

Yes, 3 write cycles, but it's still better than what I had previously -- 4 
write cycles to the internal EEPROM.  And the external writes faster than the 
internal write.

Cheers,
-Neil.




-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist